.13-micron R300 will be faster than NV30

[gf4200isdabest]

I'm sure there will be a lot of hot debates over the next few months over R300 vs NV30 vs .13-micron R300 vs NV35

The title of this topic is merely wishful thinking. What I mean by that is: I sincerely hope that the .13 micron R300 will be faster than the NV30. If it isn't, its bad for technology. This whole thing about nvidia dominating high end graphics for "since the geforce sdr" needs to be interrupted to encourage real competition. nVidia is not a company to sit idle if the R300 emerges victorious over NV30 and we'd probably see spectacular things from NV35 in that scenario. On the other hand, if NV30 just dominates the .13 micron R300, i think we'll be in for at least another year of "blah" in terms of improvements in 3d graphics.

just think along the lines of AMD vs Intel. Intel gets a kick in the face when AMD hits 1GHZ first. Intel fights back up to 2GHZ. By p4 they got too comfortable and the p4's sucked. AMD gives them another boot to the face with Athlon XP's. Intel counters by producing these sweet Northwoods of which I am one of many happy owners of. If Athlon XP's did not kick the p4's ass, I would have never had this Northwood in my comp right now because they would have been released later and been more expensive...

So here's to hoping that ATI pulls of something that I don't think anyone belives they can do: Beat nVidia through and through for an entire holiday season by getting R300 out early and then countering NV30 with .13 micron goodness....

 

[Soulkeeper]

yup
fight fight fight !!!

 

[pm]

just think along the lines of AMD vs Intel. Intel gets a kick in the face when AMD hits 1GHZ first. Intel fights back up to 2GHZ. By p4 they got too comfortable and the p4's sucked. AMD gives them another boot to the face with Athlon XP's. Intel counters by producing these sweet Northwoods of which I am one of many happy owners of. If Athlon XP's did not kick the p4's ass, I would have never had this Northwood in my comp right now because they would have been released later and been more expensive...

The design cycle for a CPU is on the order about 2-5 years. So Northwood was in development with a team of engineers working actively to design it well before the Athlon XP was announced, let alone released. During this time, there is very little one can do to substantially speed the process along, and the further along in the design process you are, the fewer changes you can make (without throwing the schedule out the window, anyway). Throwing more engineers, faster computers, etc. usually has diminishing returns - and, occassionally, can actually slow things down (training people, network problems with the new computers, etc.) In fact most CPU's "tape-out" (complete the design phase) about a year before they are actually available on the market, and sample soon after that.

My point is that the scenario you presented - of companies "kicking" other companies and then these other companies releasing new products to "fight back" isn't the way it works. The design cycles are far too long. When projects take half a decade to complete, and you have little idea what the other guys are up to, and there is very limited capability to make changes, you simply design the CPU as best as you can hoping that you will have created a product that will be competitive in the marketplace.

Back on the original subject, the shrink from 0.15um to 0.13um won't result in much of a performance improvement in itself. Usually a shrink is 30% and results in approximately a 30% speed up in logic, but TSMC and UMC broke up the shrink from 0.18um to 0.15um, and then 0.15um to 0.13um. So at best, you could hope for a 15% performance improvement. In addition power is a real problem on 0.13um, and power already is a problem for ATI with the R300, so the design cycle may take longer than expected due to problems with leakage power on the 0.13um design.

Patrick Mahoney
Microprocessor Design Engineer
Intel Corp.
Fort Collins, CO

 

[BD231]

lol, people were saying NV30was going to r*pe the R300, yet they never came to think about the die shrink. The R300 is truely amazing, and anyone saying nvidia will leapfrog ATI just like they are doing to nVidia is just plain nuts. If ATI wants to stay in this game "which they do", they will do ANYTHING it takes to stay the top dog.

 

[Nemesis77]

It's possible. NV30 will be faster than R300. It will not massacre R300 like R300 massacres GF4. But it will be faster. But that might change with the .13 micron R300. What could we expect from the refreshed R300? Maybe something like this:

350 GPU-clock (up from 325MHz. I'm being careful with my extimates)
Switch to DDR-II
Architechture tweaks (maybe additional texture-unit in the pixel-pipelines?).

Those three should boost the performance quite nicely.

 

[gf4200isdabest]

pm; i suppose you are right about it not being completely analogous because of product cycles being longer but I think the point of the analogy stands: Northwood would have been released later and been pricier without killer competition from Athlon XP's...

 

[Leo V]

Patrick, I have no doubts in what you said. I just remember how much NVIDIA improved GeForce2GTS over GeForce1 (allegedly owing to the .22--->.18u die shrink). The clock was raised from 120 to 200MHz, with twice the texturing pipes, and power consumption simultaneously went down from ~15watts to ~8watts AFAIK. There could be other explanations for this improvement (maybe better optimized design?) but ATI should be in the same position to make such improvements--they too are designing a 2nd revision of a new chip (R300). Then again, NVIDIA's engineering team is frightening--the GeForce4 uses the same .15 process as GF3, and they eked out another near-twofold performance improvement.

Your point about the long design cycle accounts for difficult-to-change release dates, but competition can still bring about lower market prices for existing products (as cumulative supply increases). What's certain, though, is that starting a thread with wishful thoughts on Anandtech forums won't bring about faster progress in technology, or fiercer competition!

Also, Intel could have intentionally produced a 1.6A Northwood stepping to regain the favor of hardware enthusiasts--no additional R&D needed to sell a chip at a lower speed/cost, right? NVIDIA did exactly this when they produced the "underclocked" Ti4200 chip to compete with ATI better.

I guess my point is that the "average customer" does benefit a lot from competition--not from faster products, but from slightly slower--and much much cheaper--ones, such as the highly popular Ti4200 and the cheaper P4's. Without competition from the "underdogs", the cheapest GF4 would be the 4400 and it would cost >$250. I also wonder if Intel would have otherwise released the lower Northwood steppings, and certainly they would have cost more.

 

[SexyK]

I think it's pretty foolish to believe that what ATI is doing will have any immidiate effects on nVidia's actions, per pm's comments. I also think it's extremely unfair to nVidia to think that they would sit on their hands if there was no competition from ATI. After 3dfx went down, nVidia pushed forward with their extremely aggresive six month product cycle despite not having a competitor in sight. I'm not saying that competition isn't good, because it's great, but I think competition has more effect on price than it does on technology. And for the record, Intel has been fairly consistent in the pricing for a long time, well before AMD was in the picture. The top end CPU always debutes ~$600 and they scale down appropriately. Only on the lower end CPUs has Intel really been forced to slash prices to offer something similar to AMD.

Kramer

 

[Aquaman]

nm

Cheers,
Aquaman

 

[Leokor]

Originally posted by: gf4200isdabest
pm; i suppose you are right about it not being completely analogous because of product cycles being longer but I think the point of the analogy stands: Northwood would have been released later and been pricier without killer competition from Athlon XP's...

In a sense, you're right. The competition from AMD did change Intel's plans about Pentium 4. But not in the way you think. On the contrary, if not for the competition, Intel would have never released the Willamette. It was a premature version of Pentium 4, while Northwood was what Intel had in mind for the first release of Pentium 4. However, Intel decided to release Pentium 4 prematurely, before the Northwood (proper Pentium 4) design was complete. Otherwise, Athlon would not have had any competition for a long time.

Leo